Early Diagnosis of Parkinson’s Disease based on Handwritten Patterns using Deep Learning

Parkinson's disease is a progressive neurodegenerative disorder. Although no cure has been found, medication can improve the symptoms. As such, early detection of the disease is essential to its treatment. However, most clinical methods used for diagnosis do not offer enough accuracy. In this work, a Histogram of Oriented Gradients is combined with a Convolutional Neural Network (CNN) to automatically detect Parkinson's disease based on handwritten patterns. The model achieved an accuracy of 87% and an F1 score of 83.21%, which outperforms state of the art clinical diagnosis techniques (79.6% accuracy without follow up, 83% with follow up).